Transmit timeouts in v1.06 (v1.09l is worse??)

[Scott Tyson]

Well I have no idea if my system is under "Heavy Load" or not.  I run
Quake2/Quake3 and some low hit HTTP/FTP. I average 30 to 50 clients.
I'm not exactly the bandwidth usage but with 50 clients I'm chewing up
most of a T1.  My ifconfig reports no errors at all, only some
collisions.  The card runs great.   My ISP uses the NICS exclusively in
al their FreeBSD servers.  Don't know if the driver code is different or
not.
Here is my info for reference:
System:
Dual PII400
256 MB RAM
EtherExpress Pro/100+
Redhat 6.0
Kernel 2.2.12 (SMP)
eeproo.c version 1.06

[scott@rand scott]$ ifconfig eth0
eth0    Link encap:Ethernet  HWaddr 00:A0:C9:E8:09:AB
          inet addr:209.63.173.4  Bcast:209.63.173.255
Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:162810831 errors:0 dropped:0 overruns:0 frame:0
          TX packets:44347630 errors:0 dropped:0 overruns:0 carrier:0
          collisions:2878207 txqueuelen:100
          Interrupt:17 Base address:0xe400

[scott@rand scott]# uptime
  8:59am  up 4 days, 16:22  (reboot for 2.2.12 kernel)

[scott@rand scott]$ more /etc/conf.modules
alias scsi_hostadapter aic7xxx
alias eth0 eepro100

I hope this helps.   If anyone has any tuning infonmation I'd be
interested.

----- Original Message -----
From: Mark Hagger <mhagger@dera.gov.uk>
To: <linux-eepro100@beowulf.gsfc.nasa.gov>
Sent: Tuesday, September 07, 1999 5:22 AM
Subject: Transmit timeouts in v1.06 (v1.09l is worse??)


> Hi,
>
> I've been running kernel 2.2.5 under Redhat 5.2 on a parallel cluster
of
> machines.  Unfortunately under high network, CPU and disk I/O load I
kept
> getting repeated Transmit timeouts messages from the eepro100 driver,
(v1.06),
> these effectively left the machine hung up and I typically had to
power off to
> reboot it.
>
> I've tried replacing v1.06 with the latest version v1.09l, but if
anything this
> was worse, under the same conditions of load the machines now fatally
crashes,
> I got a kernel oops once but it didn't appear in the syslog so I
wasn't able to
> process it.  Other than that the machine typically locks solid (blank
screen as
> well sadly), and I couldn't do anything except power off.
>
> Is anyone out there having any sucess with these eepro100 cards, I see
a number
> of people getting similiar problems with machines with high network
traffic.
>
> Unfortunately this is somewhat killing my parallel application, as I
update my
> code to get better network throughput I am able to crash/hang the
machines
> quicker......
>
> Mark